What Toolcaise can see, what it can stop, and where it stops
How Toolcaise handles agent credentials and telemetry, what Toolcaise Connect enforces on your machine, and where every control stops.
Your execution environment
Your agent continues running on infrastructure you operate. Reporting uses outbound HTTPS; no inbound port or model-provider key is required for the supervision integration.
Scoped access
Register an agent in your workspace and use an agent-scoped credential for reporting. Keep credentials in secure runner configuration and revoke compromised credentials. Account and workspace access remain separate from the runner’s reporting identity.
Metadata you choose to report
Send the operational information needed for supervision: run state, step context, timing, and optional usage or outcome details. Do not put secrets, raw prompts, or unnecessary personal information into telemetry fields.
Enforcement that stays on your machine
A Toolcaise Connect supervised run enforces its limits locally: the gateway listens only on 127.0.0.1, forwards only to the provider endpoints configured on that machine, and passes your provider key straight through without storing it. The dashboard can tighten a run’s limits but never loosen them, and can never change where model traffic, and your key, is sent.
Human decisions with explicit limits
Review and approval records are separate from agent-reported execution. Guarded actions bind authorization to an action digest and consume it once. SDK controls are cooperative and need compatible checkpoints.
Service providers
The service runs on Railway, accounts use Clerk, and payments use Stripe. The privacy policy lists every provider that handles personal information and what each one receives.
Security questions deserve specific answers
Do not infer a compliance certification, contractual SLA, guaranteed data residency, or regulatory suitability from this page. Discuss your requirements and the applicable agreement before sending regulated or sensitive data.
Know which controls are enforced and which are cooperative
Only a Connect supervised run or the MCP tool gateway sits on the agent’s path. Every other path observes, or relies on the agent’s code to comply.
| Path | What Toolcaise sees | What it can enforce |
|---|---|---|
| Python and Node.js SDKs, HTTP | What the agent reports: runs, steps, usage, outcomes | Pause, resume, cancel and exact-action approvals, at checkpoints your code implements. Cooperative: code that skips the checkpoint is not stopped. |
| OpenTelemetry | Runs, model and tool spans, and token counts from your framework’s own tracing | Nothing. It observes only. |
| Connect: Claude Code and Codex adapters | The allow-listed events and metrics those tools emit | Nothing. It reports telemetry only. |
| Connect: supervised run | Every model call made through an OpenAI- or Anthropic-compatible API, measured on your machine, plus the process’s start, end and resources | Runtime, cost, token and model-call limits and a model allowlist, by stopping the whole process tree. Pause, resume, stop and limits from the dashboard. |
| Connect: MCP tool gateway | Every tool call that passes through the proxy | Allow or deny by tool name, a call budget per session, and a hold until a person approves the exact call (Connect 0.3 or later). The approver sees the tool name; the arguments stay on your machine. |
Connect your first agent
Create a free workspace, connect the agent you already run, and watch its next run arrive. Free covers 3 agents, with no card required.